An exhaustive evaluation of modeling ecological niches above species level to predict marine biological invasions

Identifying the areas of the world with suitable environmental conditions for the establishment of invasive species represents a fundamental basis for preventing their impacts. One of the most widely used tools for this is ecological niche modeling. Nonetheless, this approach may underestimate the specie's physiological tolerances (it's potential niche) since wildlife populations of species usually do not occupy their entire environmental tolerance. Recently, it has been suggested that incorporating occurrences of phylogenetically related species improves the prediction of biological invasions. However, the reproducibility of this technique remains unclear. Here, we evaluated the generality of this protocol by assessing whether the construction of modeling units above species level improves the capacity of niche models to predict the distribution of 26 target marine invasive species. For each, we constructed supraspecific modeling units based on published phylogenies by grouping the native occurrence records of each invasive species with the records of its phylogenetically closest relative. We also considered units at species level, including only the presence of records in the native areas of the target species. We generated ecological niche models for each unit with three modeling methods (minimum volume ellipsoids - MVE, machine learning algorithms - Maxent and a presence-absence method - GLM). In addition, we grouped the 26 target species based on whether or not the species are in environmental pseudo-equilibrium (i.e., it occupies all habitats where it can disperse) and have any geographical or biological constraints. Our results suggest that the construction of supraspecific units improves the predictive capacity of correlative models to estimate the invasion area of our target species. This modeling approach consistently generated models with a higher predictive ability for species in non-environmental pseudo-equilibrium and with geographical constraints.